Upload
jade-atkins
View
215
Download
0
Tags:
Embed Size (px)
Citation preview
OLAP tool for comparing time-based data
14th May 2008
Proposed By
Pimolmas Ponchaisakuldee 104915
Advisor Dr. Paul Janecek
Contents
Introduction1
Implementation2
Experimental Evaluation3
Conclusion4
What is Data warehouse & OLAP
Data warehouse a very large database with special characteristic Data for analyzing
OLAP tool (On-Line Analytical Processing) A tool on top of data warehouse For data exploring and analyzing
3
OLAP tool Problems
Cube navigationData presentation
4
OLAP tool Problems
Cube navigation Problems of window-explorer-like tree browser
• Lose in tree topology
• Need several clicks Requirements
• Overall topology
• Multiple focuses and topology at the same time Solutions
• DOITree browser (better topology representation, multiple focuses, and go more than one level of hierarchy per click)
• DOITree VS Window-Explorer-like browser
5
OLAP tool Problems
Data presentation Tasks
• Trend finding over time• Trend comparison between 2 period of time
Problems• Pivot table: hard to do the analysis task• Existing graph: complex, hard to focus• Polaris table
– does not support hierarchy exploring of data– Problem of composition for comparison task
Solution• Focus on demand graph = graph matrix + overlaid graph
6
Objectives
To build OLAP tool To evaluate effectiveness of DOITree
visualization for OLAP-specific navigation compare to Window-Explorer-like visualization
To evaluate visual tools for time-based comparison task• Graph matrix• Overlaid graph
7
Focus on demand graph
Contents
Introduction1
Implementation2
Experimental Evaluation3
Conclusion4
Implementation
Java to modified JRubik
9
Cube Navigator
Pivot table presentation
Graph matrix presentation
Overlaid Graph
Items ListHierarchy List
Query Reset
Hierarchies on column
Items ListHierarchy List
Items on column Hierarchies on row Items on row
Design when measure is on column
10
Design when measure is on row
11
Contents
Introduction1
Implementation2
Experimental Evaluation3
Conclusion4
Evaluation 1: (WD) VS (DOI)
Window-Explorer Like browser (WD) of JRubik
DOI tree browser (DOI) of the modified JRubik
13
14
Evaluation 2: (GE) VS (FOCUS)
Overlaid Graph
General graph (GE)Focus on
demand graph (FOCUS)Graph Matrix
+VS
Experiment design
Participants will be trained to use the 2 tools The modified JRubik = (DOI) and (FOCUS) JRubik = (WD) and (GE)
Experiment design with counterbalancing order
15
Participants First Tool Second Tool Data
Student01 to Student10
The modified JRubik JRubik Foodmart
Student11 to Student20
JRubik The modified JRubik Foodmart
Staff1 and 2 The modified JRubik JRubik AIT admission
Staff3 and 4 JRubik The modified JRubik AIT admission
First Evaluation: (WD) VS (DOI)
Tasks for browser comparison Only browsers (one attribute) Appendix C page 83
• Task1 - First Node Finding • Task2 - First time Node Finding• Task3 - Subtree Revisiting• Task4 - Node Revisiting
The browsers as query tool (many attributes) Appendix D page 84
• Task1 - First time Node Finding• Task2 - Subtree Revisiting• Task3 - Node Revisiting
16
WD
DOI
WD
DOI
Second Evaluation: (GE) VS (FOCUS)
Tasks for graph comparison Simple Analysis
• Trend finding, Trend comparing Complex Analysis
• Trend comparing
17
Second Experiment: (G) VS (F)
Simple Analysis
Complex Analysis
18
Experiment Result
Browser comparison result: browsers only Subtree revisiting – users can leave subtrees open
22
First time node
finding
First node
finding
Subtree revisiting
Node revisiting
First time node
finding
First node
finding
Subtree revisiting
Node revisiting
Sig. DOI
Sig. WD
WD
DOI
Sig. WD
Sig. DOI
DOI
WD
Experiment Result
Browser comparison result: browsers only Finding for DOITree
• Advantage– More information help users to recover from mistake
• Disadvantage– More information can distract users to go wrong path
23
Experiment ResultBrowser comparison result: browser as a query
tool
24
First node finding
First node finding
Subtree revisitingSubtree revisiting
Node revisiting
Node revisiting
WD
DOI
Sig. DOI
Sig. WD
WD
DOI
Experiment ResultBrowser comparison result: browser as a query tool
(DOI) always take less clicks Go more than one level per click using (DOI) Users shrink subtree before finding new node using
(WE)
25
First node finding
First node finding
Subtree revisitingSubtree revisiting
Node revisiting
Node revisiting
25
WD
DOI
Sig. DOI
Sig. WD
WD
DOI
Experiment Result
Graph comparison result
27
FOCUSGE
Sig. FOCUS
Sig. GE
GE
FOCUS
Discussion
Personal Evaluation
32
Tasks Subtasks JPivot FreeAnalysis
JRubik Modified JRubik
T4: Analyzing visualization presentation
T4.1: How easy to read text of each graph + +++ + +++T4.2: How easy to read exact value in graph + + +++ ++++
T4.3: How easy to find trend of data + ++ + ++++T4.4: How easy to compare trend of data ++ + ++ ++++
Contents
Introduction1
Implementation2
Experimental Evaluation3
Conclusion4
Conclusion Problems
Cube navigation, Data Presentation Objectives
to evaluate effectiveness of (DOI) compare to (WD) to evaluate the focus on demand graph for time-based data analysis tasks
Browser comparison result:
34
Tasks The better Sig. Note
Browsers only
First time node finding (DOI) -
First node finding = -
Subtree revisiting (WE) Leave subtree open
Node revisiting (DOI) Some users shrink subtree
Browsers as query tool
First node finding (DOI) At middle depth (very deep has animation)
Subtree revisiting (DOI) When subtree expanded
Node revisiting (DOI) When subtree expanded
WD
DOI
Conclusion Graph comparison result:
Implications Browser comparison:
• Browser only: (DOI) is better for long term usage, (WE) is good for revisiting tasks
• Query tool: DOITree is better for unknown data and suitable for the analyzers who are always analyze unknown data
Graph comparison:• separated graph in graph matrix can eliminate complexity• Users can focus what they want• line graph is a suitable presentation when time is on X axis
35
Criteria The better Sig. Note
Pair-sample T-Test
Simple Analysis Tasks (FOCUS) -
Complex Analysis Tasks (FOCUS) -
FOCUSGE